CLIPS-LSR Experiments at TRECVID 2006

This paper presents the systems used by CLIPSIMAG and LSR-IMAG laboratories for their participation to TRECVID 2006 and the obtained results. Shot boundary detection was performed using a system based on image difference with motion compensation and direct dissolve detection. This system gives control of the silence to noise ratio over a wide range of values and for an equal value of noise and silence (or recall and precision), the F1 value is 0.805 for all types of transitions, 0.833 for cuts and 0.727 for gradual transitions. High level feature detection was performed using networks of SVM classifiers arranged in a variety of architectures and taking into account a variety of low level descriptors combining text, local and global information as well as conceptual context. The inferred average precision of our first run is 0.088. The search system uses a user controlled combination of five mechanisms: keywords, similarity to example images, semantic categories, similarity to already identified positive images, and temporal closeness to already identified positive images. The mean average precision of the system (with the most experienced user) is 0.184. 1 Shot Boundary Detection The CLIPS-IMAG team have participated to the Shot Boundary Detection (SBD) task with little modifications from previous participations. The system detects “cut” transitions by direct image comparison after motion compensation and “dissolve” transitions by comparing the norms of the first and second temporal derivatives of the images. It also contains a module for detecting photographic flashes and filtering them out as erroneous cuts and a module for detecting additional cuts via a motion peak detector. The precision versus recall or noise versus silence tradeoff is controlled by a global parameter that modifies in a coordinated manner the system internal thresholds. The system is organized according to a (software) dataflow approach and Figure 1 shows its architecture. Very little modification was made relatively to the previous versions of the system, only minor adjustments of control parameter. 1.1 Cut detection by Image Comparison after Motion Compensation This system was originally designed to evaluate the interest of using image comparison with motion compensation for video segmentation. It has been complemented afterward with a photographic flash detector and a dissolve detector. 1.1.1 Image Difference with Motion Com-